A Multi-Outcome Prognostic Model for Community-Dwelling Older Adults Admitted to Skilled Nursing Facilities for Post-Acute Care

Objectives

Nearly 20% of hospitalized older adults are discharged to a skilled nursing facility (SNF) for short-term rehabilitation. Many subsequently experience adverse outcomes, such as hospital readmissions, transitioning to long-term care rather than returning home, or death. To guide shared decision making, we developed a prognostic model for multiple outcomes for older adults admitted to SNFs.

Design

Retrospective cohort study.

Setting and Participants

Twenty percent national Medicare sample of community-dwelling older adults aged ≥66 discharged to an SNF after a hospitalization between 2017 and 2019.

Methods

We predicted 2 outcomes: 6-month all-cause mortality and “successful community discharge” (discharge to the community without subsequent rehospitalization or death within 30 days). Model predictors were pre-specified as age, sex, Elixhauser comorbidity score, hospital length of stay, elective vs urgent/emergency hospitalization, Medicaid status, principal hospital discharge diagnosis, surgical procedures, and hospitalizations in the past year. We used LASSO to reduce the 38 Elixhauser comorbidities to 12 comorbidities and logistic regression to determine separate predictor coefficients for the mortality and successful discharge outcomes. Model performance was assessed by discrimination [concordance statistic (c-statistic)] and calibration (calibration plots). Internal validation was performed via bootstrapping.

Results

The cohort included 523,740 individuals (median age 81, 62% female, 8% Black). Overall, 22% died by 6 months and 54% experienced a successful community discharge. Adjusted odds ratios varied based on outcome (eg, hospital length of stay was a stronger predictor of community discharge than mortality). The optimism-corrected c-statistics for the final model were 0.753 (95% CI, 0.752–0.755) for 6-month mortality and 0.692 (95% CI, 0.691–0.694) for successful community discharge. Calibration plots showed that the model was well calibrated for both outcomes.

Conclusions and Implications

In a national sample of older adults, this multi-outcome SNF prognostic model showed good discrimination and calibration. Risk predictions can help guide shared decision making and future planning among SNF clinicians, patients, and caregivers.